Merge remote-tracking branch 'upstream/master'

.github/workflows/Documentation.yml

@@ -12,7 +12,7 @@ jobs:
  build:
  runs-on: ubuntu-latest
  steps:
- - uses: actions/checkout@v1.0.0
+ - uses: actions/checkout@v2
  - name: Cache artifacts
  uses: actions/cache@v1
  env:

.github/workflows/juliapackage.yml

@@ -1,6 +1,8 @@
 name: Unit test
 
 on:
+ create:
+ tags:
  push:
  branches:
  - master
@@ -10,29 +12,39 @@ jobs:
  test:
  runs-on: ${{ matrix.os }}
  strategy:
+ fail-fast: false
  matrix:
- julia-version: ['1.0', '1', 'nightly']
- os: [ubuntu-latest, windows-latest, macOS-latest]
+ julia-version: ['1.6', '1', 'nightly']
+ os: [ubuntu-latest]
+ arch: [x64]
+ include:
+ - os: windows-latest
+ julia-version: '1'
+ arch: x64
+ - os: macOS-latest
+ julia-version: '1'
+ arch: x64
+
 
  steps:
- - uses: actions/checkout@v1.0.0
+ - uses: actions/checkout@v2
  - name: "Set up Julia"
  uses: julia-actions/setup-julia@v1
  with:
  version: ${{ matrix.julia-version }}
+ arch: ${{ matrix.arch }}
 
  - name: Cache artifacts
  uses: actions/cache@v1
  env:
  cache-name: cache-artifacts
  with:
- path: ~/.julia/artifacts
+ path: ~/.julia/artifacts 
  key: ${{ runner.os }}-test-${{ env.cache-name }}-${{ hashFiles('**/Project.toml') }}
  restore-keys: |
  ${{ runner.os }}-test-${{ env.cache-name }}-
  ${{ runner.os }}-test-
  ${{ runner.os }}-
-
  - name: "Unit Test"
  uses: julia-actions/julia-runtest@master
 

Project.toml

@@ -4,28 +4,21 @@ authors = ["Johnny Chen <[email protected]>"]
 version = "0.1.2"
 
 [deps]
-ColorVectorSpace = "c3611d14-8923-5661-9e6a-0046d554d3a4"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
 ImageCore = "a09fc81d-aa75-5fe9-8630-4744c3626534"
-ImageFiltering = "6a3955dd-da59-5b1f-98d4-e7296123deb5"
-MappedArrays = "dbb5928d-eab1-5f90-85c2-b9b0edb7c900"
-OffsetArrays = "6fe1bfb0-de20-5000-8ca7-80f57d26f881"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 Reexport = "189a3867-3050-52da-a836-e630ba90ab69"
 UUIDs = "cf7118a7-6976-5b1a-9a39-7adc72f591a4"
 
 [compat]
-ColorVectorSpace = "0.7, 0.8, 0.9"
-Distributions = "0.19, 0.20, 0.21, 0.22, 0.23, 0.24"
-ImageCore = "0.8.4"
-ImageFiltering = "0.5, 0.6"
-MappedArrays = "0.2, 0.3"
-OffsetArrays = "0.10, 0.11, 1"
+Distributions = "0.19, 0.20, 0.21, 0.22, 0.23, 0.24, 0.25"
+ImageCore = "0.9"
 Reexport = "0.2, 1.0"
-julia = "1"
+julia = "1.6"
 
 [extras]
 BM3DDenoise = "95fb3b36-088a-43fb-bb1b-b1f34fadbd7d"
+ImageIO = "82e4d734-157c-48bb-816b-45c225c6df19"
 ImageMagick = "6218d12a-5da1-5696-b52f-db25d2ecc6d1"
 ImageQualityIndexes = "2996bd0c-7a13-11e9-2da2-2f5ce47296a9"
 ImageTransformations = "02fcd773-0e25-5acc-982a-7f6622650795"
@@ -34,4 +27,4 @@ Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 TestImages = "5e47fb64-e119-507b-a336-dd2b206d9990"
 
 [targets]
-test = ["BM3DDenoise", "ImageMagick", "ImageQualityIndexes", "ImageTransformations", "ReferenceTests", "Test", "TestImages"]
+test = ["BM3DDenoise", "ImageIO", "ImageMagick", "ImageQualityIndexes", "ImageTransformations", "ReferenceTests", "Test", "TestImages"]

docs/Project.toml

@@ -5,13 +5,15 @@ Documenter = "e30172f5-a6a5-5a46-863b-614d45cd2de4"
 FileIO = "5789e2e9-d7fb-5bc7-8068-2c6fae9b9549"
 ImageCore = "a09fc81d-aa75-5fe9-8630-4744c3626534"
 ImageDistances = "51556ac3-7006-55f5-8cb3-34580c88182d"
+ImageIO = "82e4d734-157c-48bb-816b-45c225c6df19"
 ImageMagick = "6218d12a-5da1-5696-b52f-db25d2ecc6d1"
 ImageNoise = "2e361c58-64d7-11e9-3ab3-89c4c0b8f7db"
 ImageQualityIndexes = "2996bd0c-7a13-11e9-2da2-2f5ce47296a9"
 ImageShow = "4e3cecfd-b093-5904-9786-8bbb286a6a31"
+ImageTransformations = "02fcd773-0e25-5acc-982a-7f6622650795"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 TestImages = "5e47fb64-e119-507b-a336-dd2b206d9990"
 
 [compat]
-Documenter = "0.24"
-DemoCards = "0.2"
+DemoCards = "0.4"
+Documenter = "0.27"

docs/examples/apply_noise/AWGN.jl

@@ -1,18 +1,22 @@
 # ---
 # title: White Additive Gaussian Noise
+# id: wagn-demo
 # cover: assets/awgn_cover.png
-# description: This demo shows how to add white additive gaussian noise to an image
+# description: Add white additive gaussian noise to an image
+# date: 2019-12-22
+# author: Johnny Chen
 # ---
 
 # Mathematically, adding a white additive gaussian noise to an image is as simple
 # as $Y = X + N$ where $N \sim \mathcal{N}(\mu,\,\sigma^{2})\,.$
 
-using ImageNoise, TestImages, ImageShow, ImageCore, ImageQualityIndexes
+using ImageNoise
+using TestImages, ImageShow, ImageCore, ImageQualityIndexes, ImageTransformations
 using FileIO, Random #src
 
 # Let's add noise to gray image first
 
-gray_img = testimage("cameraman") .|> float32
+gray_img = float.(imresize(testimage("cameraman"), ratio=0.5))
 noise = AdditiveWhiteGaussianNoise(0.1)
 Random.seed!(0) #src
 gray_noisy_img = apply_noise(gray_img, noise)
@@ -22,7 +26,7 @@ gray_noisy_img = apply_noise(gray_img, noise)
 # Adding noise channel by channel would indeed get a different distribution (but
 # still a gaussian noise).
 
-rgb_img = testimage("mandrill") .|> float32
+rgb_img = float.(imresize(testimage("mandrill"), ratio=0.5))
 Random.seed!(0) #src
 rgb_noisy_img = apply_noise(rgb_img, noise)
 
@@ -35,7 +39,8 @@ hsv_noisy_img = apply_noise(hsv_img, noise)
 incorrect_hsv_noisy_img = colorview(HSV, apply_noise(channelview(hsv_img), noise))
 
 ## If we compare their psnr, the second one is significantly lower than the first one
-[psnr(rgb_noisy_img, RGB.(hsv_noisy_img))
-psnr(rgb_noisy_img, RGB.(incorrect_hsv_noisy_img))]
+[assess_psnr(rgb_noisy_img, RGB.(hsv_noisy_img))
+assess_psnr(rgb_noisy_img, RGB.(incorrect_hsv_noisy_img))]
 
-save("assets/awgn_cover.png", rgb_noisy_img) #src
+mkpath("assets") #src
+save("assets/awgn_cover.png", rgb_noisy_img) #src

docs/examples/reduce_noise/nonlocal.jl

@@ -1,24 +1,32 @@
 # ---
 # title: Non-local Mean Filter
+# id: nlmf-demo
 # cover: assets/nlmean_cover.png
-# description: This demo shows how to use Non-local Mean filter to reduce gaussian noise
+# date: 2019-12-22
+# author: Johnny Chen
+# description: Use Non-local Mean filter to reduce gaussian noise
 # ---
 
-using ImageNoise, TestImages, ImageShow, ImageCore, ImageQualityIndexes
+using ImageNoise
+using TestImages, ImageShow, ImageCore, ImageQualityIndexes, ImageTransformations
 using FileIO, Random #src
 
+# First, load an image and and add gaussian noise to it
 
-gray_img = testimage("cameraman") .|> float32
+gray_img = float.(imresize(testimage("cameraman"), ratio=0.5))
 n = AdditiveWhiteGaussianNoise(0.1)
 noisy_img = apply_noise(gray_img, n)
 
+# Then calling the standard `reduce_noise` API
+
 f_nlmean = NonlocalMean(0.1)
 denoised_img = reduce_noise(noisy_img, f_nlmean)
 
-[gray_img noisy_img denoised_img]
+mosaicview(gray_img, noisy_img, denoised_img; nrow=1)
 
-#+
+# Get the PSNR using ImageQualityIndexes package:
 
-psnr(gray_img, denoised_img)
+assess_psnr(gray_img, denoised_img)
 
-save("assets/nlmean_cover.png", denoised_img) #src
+mkpath("assets") #src
+save("assets/nlmean_cover.png", denoised_img) #src

src/ApplyNoise/ApplyNoise.jl

@@ -1,10 +1,9 @@
 module ApplyNoise
 
 using Random
-using MappedArrays
 using ImageCore
+using ImageCore.MappedArrays
 using ImageCore: NumberLike, GenericGrayImage, GenericImage
-using ColorVectorSpace
 
 import ..NoiseAPI: AbstractImageNoise, apply_noise, apply_noise!
 import Base: show

src/ReduceNoise/NonlocalMean.jl

@@ -81,7 +81,14 @@ function (f::NonlocalMean)(out::AbstractArray{<:NumberLike, 2},
  R = CartesianIndices(img) # indices without padding
 
  oₚ = ntuple(_->r_p, ndims(img))
- img = padarray(img, Pad(:symmetric, oₚ, oₚ))
+ padded_axes = map(axes(img), oₚ) do ax, o
+ first(ax)-o:last(ax)+o
+ end
+ img = PaddedView(zero(eltype(img)), img, padded_axes)
+ # PaddedView's `getindex` has some overhead so we collect it into a dense array
+ # and then pass to OffsetArray. Becuase we're doing a heavy amount of `getindex`
+ # operation, this extra memory allocation worths.
+ img = OffsetArray(collect(img), padded_axes)
  Δₚ = CartesianIndex(oₚ)
  Δₛ = CartesianIndex(ntuple(_->r_s, ndims(img)))
 
@@ -137,23 +144,23 @@ end
 
 """ gaussian-like kernel """
 function make_kernel(T, r)
- kernel = centered(zeros(2r+1, 2r+1))
+ kernel = OffsetArrays.centered(zeros(2r+1, 2r+1))
  R = CartesianIndices(kernel)
  for d in 1:r
  v = 1/(2d+1)^2
  for i in R[-d:d, -d:d]
  kernel[i] += v
  end
  end
- centered(T.(kernel ./ sum(kernel)))
+ OffsetArrays.centered(T.(kernel ./ sum(kernel)))
 end
 
 """ weighted squared euclidean """
 function wsqeucliean(W, X, Y)
  rst = zero(eltype(W))
  # use linear indexing
  @inbounds @simd for i = 1:length(W)
- rst += W[i] * abs2(X[i] - Y[i])
+ rst += W[i] * _abs2(X[i] - Y[i])
  end
  return rst
 end

src/ReduceNoise/ReduceNoise.jl

@@ -1,12 +1,11 @@
 module ReduceNoise
 
 using Base.Iterators
-using MappedArrays
-using OffsetArrays
 using ImageCore
+using ImageCore.MappedArrays
+using ImageCore.OffsetArrays
+using ImageCore.PaddedViews
 using ImageCore: NumberLike, GenericGrayImage, GenericImage
-using ImageFiltering
-using ColorVectorSpace
 import ..NoiseAPI: AbstractImageDenoiseAlgorithm, reduce_noise, reduce_noise!
 
 include("compat.jl")

src/ReduceNoise/compat.jl

@@ -8,3 +8,18 @@ if v"1.0" <= VERSION < v"1.1"
  _colon(I::CartesianIndex{N}, J::CartesianIndex{N}) where N =
  CartesianIndices(map((i,j) -> i:j, Tuple(I), Tuple(J)))
 end
+
+# patch for ColorVectorSpace 0.9
+# for CVS < 0.9, we can just use the fallback solution in Distances
+if isdefined(ImageCore.ColorVectorSpace, :⊙)
+ # Because how abs2 calculated in color vector space is ambiguious, abs2(::RGB) is un-defined
+ # since ColorVectorSpace 0.9
+ # https:/JuliaGraphics/ColorVectorSpace.jl/pull/131
+ @inline _abs2(c::Colorant) = mapreducec(v->float(v)^2, +, 0, c)
+ @inline _abs2(x) = abs2(x)
+ @inline _abs(c::Colorant) = mapreducec(v->abs(float(v)), +, 0, c)
+ @inline _abs(x) = abs(x)
+else
+ const _abs2 = abs2
+ const _abs = abs
+end